The world’s richest man just outdid himself, again.
Last year, Elon Musk and his xAI team turned an empty warehouse the size of 13 football fields into the largest artificial intelligence (AI) training cluster in the world in just 122 days ‘Colossus 1’.
Elon and his team are building Colossus 2, which is on track to be the new world’s largest AI data center. Colossus 2 will scale to about half a million AI chips when fully built. It will consume as much electricity as the entire city of Miami.
For the last two years, everyone in AI, obsessed over chips. Elon Musk joked that GPUs were “harder to get than drugs.”
- But the chokepoint in AI isn’t chips anymore…
It’s electricity.
The bigger the AI model, the better the results.
That’s why AI models just keep getting bigger and bigger.
And this year alone, big tech will spend $350 billion+ building new AI data centers.
In one year, AI infrastructure spend has grown to rival the entire Apollo program, when adjusted for today’s dollars.
Back in 2006, Google (GOOG) spent $600 million on its first big data center in Oregon. At the time, it was a huge deal.
Fast-forward to today. OpenAI, Oracle Corp. (ORCL), and SoftBank are spending almost 100 times as much—$500 billion—on their “Stargate” project. Stargate will hoover up enough energy to power a city the size of New York!
Yesterday’s megaprojects were highways and dams. Today’s megaprojects are digital cities of silicon and steel, fed by rivers of electricity.
- Microsoft (MSFT) added two nuclear power plants of capacity last year.
That was just to keep AI running.
Meta Platforms (META) went a step further. It’s launching its own energy trading arm to manage the power costs of its massive data center footprint.
At this scale, there’s no choice. To keep AI alive, big tech has to build, buy, or control its own power.
This is how disruption works. It redefines entire industries.
And as investors, we want to own the companies enabling this disruption ‘Utilities’.
Busy And Useful 